Boban Stojanović is an associate professor at the Department of Mathematics and Informatics, Faculty of Science, University of Kragujevac, Serbia and the founder of Vodéna, an innovative IT company focused on university-strength research, data analysis, computer modeling, and optimization, all integrated through flexible and efficient software applications.
After a number of developed scientific methods and software he sees himself somewhere between scientist and software engineer. Currently he leads a group of about 20 researchers and programmers who work on development of industrial software solutions for various real-world problems.
The focus of Q4Q is on the accuracy and effectiveness of AI in the testing of software solutions before roll-out as well as the role of historical data and human influence in AI-tested processes. However, contemporary software solutions increasingly rely on the components based on some kind of machine learning or artificial intelligence. Therefore, future software QA should involve not only testing the correctness of the written code, but also controlling the quality of the underlying ML and AI models in terms of accuracy, confidence, reliability, and efficiency.
In this talk, Boban will try to answer these questions:
How can we know what level of accuracy we can expect from AI models built from available data? How to speed up the process of building an AI model? Can we use the tools of Automated machine learning (AutoML) for rapid prototyping process?
How can we be sure that we have chosen the best of all possible AI models? Do existing solution-driven AutoML approaches meet the challenges we face or do we have to turn to tailor-made, problem-driven solutions?
What is the robustness of our AI model in case of lack or low quality of input data? Is it enough to employ static ML models or we have to focus on systems capable of adapting to ever-growing data-sets and new data sources in near-real-time?
How Evolving Deep Neural Networks can help answer these and many other questions will be heard in this talk.